arXiv cs.AI· Mingzhan Yang, Weili Wu·· 10 小时前AI 评分31
临床 LLM 在证据不完整时如何决策：DECIDE/ASK/DEFER 框架评估 Qwen、Gemini 与 GPT
Decide, Ask, or Defer: Clinical LLMs under Incomplete Evidence
AI 导读
研究提出 DECIDE/ASK/DEFER 三态框架与盲测协议，在 DDXPlus 构建的 200 个匹配临床证据状态上评估 Qwen、Gemini 和 GPT，三者在相同证据下的行动选择出现 137 个状态分歧。
正文 · 原文
Abstract:Clinical LLMs must decide not only what diagnosis to produce, but also whether the available evidence is sufficient for autonomous decision making. Binary DECIDE/ABSTAIN formulations merge distinct non decision states and do not explicitly evaluate information acquisition. We introduce a DECIDE/ASK/DEFER formulation together with a blinded protocol that prevents models from using evidence completeness metadata. We evaluate Qwen, Gemini, and GPT on 200 matched clinical evidence states constructed from DDXPlus. The models show substantial differences in action selection under identical evidence, with disagreement in 137 of 200 states. For Qwen, a matched targeted versus random analysis shows that selected information changes the likelihood of a subsequent autonomous decision more clearly than diagnostic correctness. Its matched DECIDE/ABSTAIN baseline further reveals a safety autonomy tradeoff: the three action policy rescues some erroneous autonomous decisions but also removes some correct autonomous deci sions. These results show that separating information acquisition from clinician deferral exposes behavior that binary abstention hides, without yielding a uniformly improved decision policy.
| Comments: | Accepted to the Main Track of the NeurIPS 2026 GenAI4Health Workshop |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.04542 [cs.AI] |
| (or arXiv:2610.04542v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.04542 arXiv-issued DOI via DataCite |
Submission history
From: Mingzhan Yang [view email]
[v1]
Sat, 3 Oct 2026 14:16:09 UTC (127 KB)
[v2]
Tue, 6 Oct 2026 23:26:46 UTC (127 KB)
[v3]
Mon, 9 Nov 2026 21:39:31 UTC (127 KB)
来源：arXiv cs.AI · arxiv.org